who bears the economic costs of environmental regulations? · 2017-08-16 · who bears the economic...
TRANSCRIPT
NBER WORKING PAPER SERIES
WHO BEARS THE ECONOMIC COSTS OF ENVIRONMENTAL REGULATIONS?
Don FullertonErich Muehlegger
Working Paper 23677http://www.nber.org/papers/w23677
NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts Avenue
Cambridge, MA 02138August 2017
For helpful comments and suggestions, we would like to thank Tatyana Deryugina, Arik Levinson, and Reed Walker. This paper is an invited submission to the Review of Environmental Economics and Policy (REEP), symposium on “Distributional Effects of Environmental Policy.” The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.¸˛
NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications.
© 2017 by Don Fullerton and Erich Muehlegger. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source.
Who Bears the Economic Costs of Environmental Regulations?Don Fullerton and Erich MuehleggerNBER Working Paper No. 23677August 2017JEL No. H22,H23,Q48,Q52
ABSTRACT
Public economics has a well-developed literature on tax incidence – the ultimate burdens from tax policy. This literature is used here to describe not only the distributional effects of environmental taxes or subsidies but also the likely incidence of non-tax regulations, energy efficiency standards, or other environmental mandates. Recent papers find that mandates can be more regressive than carbon taxes. We also describe how the distributional effects of such policies can be altered by various market conditions such as limited factor mobility, trade exposure, evasion, corruption, or imperfect competition. Finally, we review data on carbon-intensity of production and exports around the world in order to describe implications for effects of possible carbon taxation on countries with different levels of income per capita.
Don FullertonDepartment of FinanceUniversity of Illinois515 East Gregory Drive, BIF Box#30 (MC520)Champaign, IL 61820and [email protected]
Erich MuehleggerUniversity of California - DavisEconomics DepartmentOne Shields AveDavis, CA 95616and [email protected]
1. Introduction
Historically, command and control approaches have dominated environmental regulation.
Examples in the U.S. include technology standards embedded in non-attainment state
implementation plans in response to the 1970 Clean Air Act; state renewable portfolio
standards; and the recent water restrictions in California. These kinds of mandates account
for a majority of environmental regulations not only in the U.S. but also in other countries
around the world. Despite the ubiquity of environmental mandates, most research
examining the distributional consequences of environmental policies has focused on
environmental taxes or other market-based approaches, the burdens from effects on output
prices rather than on factor prices, and the effects in developed countries.
This paper has four main objectives. First, we summarize existing understanding of
tax incidence, defined as the distributional consequences of environmental taxes or
pollution pricing. Starting with some theoretical literature, we describe how producers and
consumers split the burden of environmental policy, and how such policies affect returns to
factors like capital and labor, or prices of clean and dirty inputs. For market-based
approaches, we can draw on the large existing literature summarizing these effects,
including the well-developed empirical literature on the incidence of environmental taxes.
Second, we extend these existing concepts to an initial understanding of the likely
distributional effects of mandates, as discussed for example in Fullerton and Heutel (2010)
and Goulder et al. (2016). We then review a new and emerging empirical literature on
distributional effects of non-tax environmental mandates and regulations.
A third objective of the paper is to describe how alternative market conditions can
affect these distributional results. Initial models of tax incidence are based on assumptions
of perfect competition, perfect factor mobility, and perfect enforcement – either in a closed
economy or in a small open economy with free trade at fixed world prices. We therefore
discuss how the burden of environmental taxes and mandates might vary along five
dimensions particularly relevant to developing countries: (a) the degree of labor and capital
mobility between sectors, (b) the mobility of those factors between jurisdictions, (c) the
openness of an economy to trade, (d) the degree of regulatory evasion, and (e) market
concentration in the regulated market.
Labor and capital mobility help determine the ways in which domestic wages and
rental rates depend on worldwide wage rates and rental prices, rather than just on domestic
-2-
markets. Openness to trade helps determine the extent to which goods’ prices depend on
world prices rather than domestic markets. Regulatory evasion affects the distribution of
the burden of environmental policy, and it poses greater challenges in developing countries
where monitoring and enforcement tend to be imperfect. Finally, market power can also
affect the distribution of burdens from taxes – and presumably from non-tax regulations.
Finally, we discuss how existing research ought to be able to inform the incidence
of environmental regulation in developing countries. With the exception of a small
literature, as reviewed by Greenstone and Hanna (2014) and Greenstone and Jack (2015),
the vast majority of work on the distribution of the burdens of environmental policy has
focused on the developed world. Yet, the market imperfections faced in the developed
world may differ from those faced by developing countries. Consequently, empirical
estimates of environmental incidence in the U.S. or Europe may not be a good guide to the
international experience. Furthermore, given the projected growth of the developing world,
it is in these locations where future environmental regulation may be most important.
We cannot apply all of the concepts just described to all environmental taxes
around the world, but we find it somewhat more manageable to take the specific example
of a widespread carbon tax and to describe its likely distributional effects on countries at
different levels of per capita income. To do so, we group countries into four income
categories and calculate summary statistics for each group such as their carbon-intensity of
electricity generation, the importance of their direct fossil fuel exports, and their relative
dependence on exports of carbon-intensive manufactured goods. Although low-income
countries tend to be net importers of fossil fuels (particularly refined petroleum products),
we do not find strong relationships between carbon-intensity and country income. Rather,
we find substantial variation in both carbon intensity and trade exposure, even after
conditioning on income. These results suggest that assessing the distributional impacts of
carbon policy broadly by country income may overlook important types of variation.
2. The Theory of Tax Incidence Applied to Environmental Policy
When a tax is applied to each unit of an input to production or to an output, the law might
state that the seller or that the buyer must pay the tax, but economics textbooks are clear
that this “statutory incidence” has little bearing on who really bears the burden. Actual
distributional effects are known as “economic incidence.” Here, we consider pollution as
an input to production, in order to study a pollution tax or other pollution policy that
-3-
similarly raises the cost of that input. How much of this burden is passed to consumers?
In a simple partial equilibrium model, a pollution policy will increase the cost of
production and thus raise the output price by an amount that depends on the pollution-
intensity of production. For primary energy goods like fuels, for example, carbon intensity
can be measured in carbon emissions per unit of heat generated by the fuel. For each unit
of a produced good, carbon-intensity equals the energy required at each stage of its
production or consumption times the carbon per unit of the energy source employed. Then
the burden of the extra tax cost is generally split between producers and consumers in a
way that depends on market characteristics. Under perfect competition, economic
incidence depends on the relative demand and supply elasticities of the good, where the
more inelastic side of the market bears a greater fraction of the costs of the policy.1 In
other words, if demand for fossil-fuel-intensive goods like gasoline or electricity is more
inelastic than supply, then consumers bear more of the burden than producers.
Here we make two additional points. First, the same kind of elasticity analysis can
be applied to a market for an input to production, including supply by the natural resource
owner and demand by a firm. The burden of a tax on that carbon-intensive resource can be
higher for the owner/supplier if supply is inelastic, or it can be higher for the purchasing
firm if demand is inelastic. Second, these demand elasticities depend on substitution
possibilities, which themselves depend on legal restrictions and other market conditions.
For example, the existence of unregulated sectors or different production technologies
allows firms to substitute away from regulated inputs and outputs, thereby avoiding
regulation, similar to the substitution in Gibson (2015).
The usual graduate course in public economics includes a well-developed general
equilibrium theory of tax incidence dating at least back to Harberger’s (1962) analytical
two-sector model. He studies the incidence of the corporate income tax in a model with
two sectors and two inputs – labor and capital – owned by a single aggregate household.
One sector produces a corporate manufactured output, X, and the other produces a non-
corporate output, Y (an aggregation of agriculture, services, and housing). Harberger
models the corporate income tax as a “partial factor tax” on capital used in the corporate
sector, KX. Equations are then differentiated to linearize the model and solve N equations
1 If η is the supply elasticity, and ε is the absolute value of the demand elasticity, then Fullerton and Metcalf (2002) show that ε/(ε+η) is the fraction of the price increase borne by suppliers (which rises when demand is more elastic), and η/(ε+η) is the fraction borne by consumers (which rises when supply is more elastic).
-4-
for N unknowns, including the key price changes, in terms of all model parameters and an
exogenous small change in the tax, tKX.
The incidence of the tax in this context has two major components: the “sources
side” refers to burdens from changes in the rental/wage ratio that affect sources of income,
and the “uses side” refers to burdens from changes in relative commodity prices that affect
uses of income. This partial factor tax raises the cost of production in the corporate sector,
so it raises the relative price of X, but it also raises the cost of capital and induces firms to
substitute from capital to labor. Mobile capital can bid down the return in the noncorporate
sector, but if the shrinking corporate sector is labor intensive then labor also moves and
can bid down the wage in the noncorporate sector; the overall incidence via the rental/wage
ratio depends on parameters that affect the relative strength of these two effects.
Hundreds of subsequent articles extended this model to consider unemployment,
international trade, imperfect mobility, imperfect competition, and other market conditions.
The model was extended to more sectors, or more factors, or other tax instruments.2 In
particular, Fullerton and Heutel (2007) extend Harberger’s 1962 model to consider a clean
sector and a dirty sector with inputs of labor, capital, and pollution. Three inputs to
production mean that either labor or capital can be a relative substitute or complement for
pollution, so a tax on pollution can increase relative demand for labor or for capital – with
consequent effects on the rental/wage ratio (and thus burdens on the sources side).
For at least two reasons, theorems from these Harberger-style general equilibrium
models are virtually untestable. First, the nature of a general equilibrium model means that
everything is endogenous, as the tax change can affect all prices and all quantities in the
entire economy. Second, each model is designed to consider a one-time hypothetical policy
shock in a closed economy, all else equal, in order to identify the effects of this tax alone
with no other exogenous shocks to technology, no new conflicts, nor changes in other
nations’ policies. Consequently, empirical studies get to observe precious little random
variation in the key tax rate across many jurisdictions or time periods, and they are
bedeviled by other simultaneous changes in the economy. We return below to discussion
of valiant attempts at empirical measurement.
3. Can We Use the Theory of Tax Incidence for Non-Tax Regulations?
2 For example, Mieszkowski (1972) studies the residential property tax in a similar analytical general equilibrium model with three factors of production (labor, capital, and land).
-5-
We next turn to the observation above that most actual pollution policies are not taxes at
all. How can tax incidence theory be used to analyze the burdens of non-tax environmental
technology standards? The U.S. has employed some pollution-pricing instruments since
the Acid Rain Program introduced sulfur dioxide permits in 1990, but like most other
nations, the U.S. still employs a bewildering array of emission rate standards for
automobile manufacturers, renewable portfolio standards for electricity generating
companies, and other mandates to control emissions or energy use.
An answer to that question is in the nature of those regulations. Instead of imposing
the same emission limit on all firms of different sizes, these rules tend to impose on all
firms the same maximum ratio, such as emissions per mile, carbon per kilowatt hour, or
energy use per degree of cooling. As it turns out, such ratio standards can adequately be
represented in general-equilibrium models. Fullerton and Heutel (2010) note that a newly
binding ceiling on emissions per unit output can be satisfied by reducing emissions in the
numerator, or by increasing output in the denominator, or by some of both.
Thus, as shown by them and by Goulder et al. (2016), that ratio standard is
equivalent to the revenue-neutral combination of an implicit tax on emissions and subsidy
to output. This insight is important for incidence analysis, because the emissions tax alone
would raise the cost of production and thus raise the break-even product price. Thus, the
addition of the implicit output subsidy tends to offset some of the likely increase in the
relative price of that output. For example, the limit on carbon-per-kwh in the U.S. Clean
Power Plan is effectively an implicit tax on carbon and subsidy to electricity. That subsidy
helps prevent increases in the price of electricity (and burdens on the uses side).
A carbon tax, however, has both a “substitution effect” among inputs to production
that could reduce carbon per kwh and an “output effect” that further reduces carbon by
raising the price of output and reducing the purchase of kilowatt hours. Because low-
income families devote a higher fraction of total spending to electricity than high-income
families, the carbon tax is often found to have a regressive burden.3 In contrast, a mandate
on emission rate might be less efficient than the carbon tax alone, by failing to take
advantage of the carbon-reducing output effect just mentioned. On the other hand, it might
also be less regressive, if it does not raise the price of electricity as much. Thus arises the
3 But see Cronin et al. (2017) for an alternative view. Because public transfers increase automatically by a consumption price index, low-income families are protected against the costs of climate policy. When families are ranked by annual consumption as a proxy for lifetime income, a carbon tax is progressive.
-6-
possibility of a tradeoff between economic efficiency and distributional objectives.
Because a carbon tax would raise the price of electricity and gasoline, and place
disproportionate burdens on low-income households who spend a high fraction of their
budget on those necessities, policymakers may be concerned about regressivity.4 They
might therefore prefer mandates such as CAFE standards for vehicles or energy efficiency
standards for appliances (e.g. refrigerator, furnace, or air conditioner). But, what are the
distributional effects of those mandates? 5
Most empirical work is partial equilibrium in nature. To measure incidence of
corporate average fuel economy (CAFE) standards, Davis and Knittel (2016) note that
automakers are required to sell more fuel-efficient cars and fewer inefficient cars, so
retailers reduce prices of the former and raise prices of the latter. These CAFE obligations
are tradable, so permit prices can be used to quantify the implicit subsidy and tax rates.
They have comprehensive data on all U.S. vehicle registrations in 2012, including rich data
on characteristics of those cars, and they measure CAFE burdens by changes in a family’s
vehicle costs. For new vehicles, they find that CAFE is mildly progressive (because high-
income households buy more new vehicles). Including effects on used vehicle costs makes
CAFE mildly regressive. They also compare alternative policies. They find that a carbon
tax alone is more regressive than CAFE standards, but progressivity can more easily be
achieved by a carbon tax with a revenue rebate such as through uniform transfers.
Levinson (2016) tackles a similar question but uses other data that include both
vehicle information and miles driven. He finds that energy efficiency standards are even
more regressive than a carbon tax. His logic is that without such policies, low-income
families would tend to purchase inexpensive cars (or appliances), and use them less
extensively than do high-income families. In contrast, “richer households will purchase
more energy efficiency and more energy” (p.10).
Thus, the imposition of an energy efficiency standard requires the manufacturer to
sell only the more expensive cars or appliances with greater energy efficiency, which
reduces the cost per unit of services from the appliance (e.g. heating in the winter, cooling
4 Electricity is a relatively high fraction of spending for low-income U.S. families, but gasoline is a low fraction of spending for the very poor who do not own cars. Various papers in the book edited by Thomas Sterner (2012) show that a gasoline tax would be progressive in poor countries where only rich own cars. 5 A new literature is emerging on distributional effects of mandates via invitations to a special issue of the Journal of the Association of Environmental and Resource Economists (JAERE) including working papers by Davis and Knittel (2016), Levinson (2016), Reguant (2017), and Bruegge, et al. (2016).
-7-
in the summer, or miles driven in the vehicle). Low-income households do not get to take
much advantage of those lower-cost services, since they do not want – or cannot afford –
to make extended use of those vehicles or appliances, so they effectively are just hit with
the raised cost of purchasing such equipment. In contrast, higher-income families may
have wanted to purchase the greater energy efficiency anyway, because they can afford its
higher price and want to run the equipment more extensively. The mandate then puts less
burden on high-income families.
Both of these papers show that mandates are more regressive than a carbon tax
when compared in a revenue-neutral way, and so both undercut the case for energy
efficiency standards on distributional grounds. Economists have often shown that mandates
are less cost-effective than pricing the pollutant, and now they have shown that such
mandates are also more regressive.
Other kinds of mandates may impose constraints on electricity providers. For
example, a renewable portfolio standard (RPS) may require a certain percentage of
generation from sources like wind and solar power. In order to compare efficiency and
distributional effects of such mandates to other policies like a tax on carbon or a subsidy to
renewables, Reguant (2017) constructs a partial-equilibrium simulation model with
multiple generation technologies and demand sectors.
Previous simulation studies have assumed that final consumers face wholesale
prices, which tend to rise with an RPS or carbon tax but fall with renewable subsidies.
However, Reguant explicitly accounts for retail market design, where renewable
production subsidies are often financed by retail tariffs on consumers. She finds that the
carbon tax has two efficiency advantages: first, it can induce substitution not only to the
relatively expensive renewables but also to inexpensive natural gas. Second, it incorporates
demand responses. Results indicate that the carbon tax imposes burdens on consumers but
provides gains to producers of hydro and nuclear power. Interestingly, renewable policies
are made substantially more efficient if their costs are reflected in retail prices, though of
course benefits still go to producers of renewables. Overall, these results point to the
importance of details in policy design, electricity pricing, and potential tradeoffs between
economic efficiency and distributional impacts.
Lastly, the burgeoning new literature on distributional effects of mandates includes
analysis of state or local building codes that may achieve reduced energy use by requiring
better insulation, particular building features, or overall energy efficiency. To identify the
-8-
effects of actual building energy codes, Bruegge et al. (2016) use spatial variation in the
strictness of California’s building requirements created by the state’s 16 distinct sets of
climate zone requirements. They use data on over 350,000 homes located within five
kilometers on both sides of each climate zone border, and household electricity and natural
gas billing data for the years 2009-2015 from four California utilities, to estimate the effect
of these building codes on characteristics of houses built, on energy use, and on sales
prices. They find that building codes reduce total energy consumption by about 6.2% for
the average household, primarily by reducing square footage – particularly in the top two
income deciles. Codes thus impose costs primarily on the rich, while the bottom two
deciles are essentially unaffected. But, on a per square-foot basis, no income group
experiences a significant change in house prices.
4. Other Considerations Can Alter the Incidence of Environmental Policy
In several ways, existing research has modified the basic theory described above to relax
restrictive assumptions such as perfect mobility, costless trade, perfect enforcement, and
perfect competition. These considerations are particularly important for developing
countries where market imperfections or other frictions may play large roles. Moreover,
these market conditions can change the answer about who bears the burden of regulation.
a. The Degree of Inter-sectoral Factor Mobility
The basic model discussed above is a long run general equilibrium model used to think
about output prices and factor returns after all adjustments. Workers might be displaced in
the short run, but are expected be able to find a new job in an expanding industry at the
going market wage. An even better view of the model, however, is not that it considers the
move from one equilibrium to a new equilibrium with a different tax policy, because that
would require costly adjustments. Instead, it compares two different states of the world:
one where this policy never existed and the other where this environmental policy always
existed. It is not a move from one to the other, but a conceptual comparison to isolate the
impact of having the policy versus not having the policy, all else equal. It compares two
worlds with the same technology, the same preferences, and the same factor endowments.
However, many discussions of proposed environmental regulations are often mostly
about short run adjustment costs and the imperfect mobility of workers. Coal miners live in
a town dominated by the coal industry, a town where they have lived all their life and do
-9-
not want to leave. Moreover, they have industry-specific skills that would be not be valued
when looking for a job in a different industry. The cost of displacement is not just lost
wages, as job loss can negatively affect the displaced worker’s health (see e.g., Sullivan
and Von Wachter (2009)). It can also cause severe psychological trauma.6
The lost value of this imperfectly mobile human capital is analogous to the lost
value of imperfectly mobile physical capital. If a large carbon tax were to apply to all coal
used by existing power plants, then the market value of the stock in companies that own
coal-fired power plants would fall significantly. Losses are capitalized into the stock price
of those companies, with potentially huge burdens on those who happen to own shares in
that company at the time of the policy change (see e.g., Bushnell et al. 2013). Analogously,
losses are capitalized into the market value of human capital for those who work in coal
mines and those with specific training to work in coal-fired power plants.
b. Inter-jurisdictional Factor Mobility
The discussion of the theory of tax incidence above pertains to a simple static model of a
closed economy, with a fixed amount of labor and capital, so changes in the demand for
either factor can affect its return. In that way, part of the burden of any environmental tax
or mandate might not just affect buyers of pollution-intensive products on the uses side,
but can affect long-run relative returns to workers or investors on the sources side. That
assumption might have been appropriate for a large economy like the U.S., at least in past
decades, but small countries depend on international capital markets, and even the U.S. is
subject to increasing globalization. Furthermore, as Fowlie (2009) highlights for a large
country like the U.S., the presence of unregulated sectors create analogous effects.
Before looking at the degree of factor mobility, first consider the opposite extreme.
Instead of a fixed stock of capital or vertical supply curve as in the Harberger model,
consider the case of perfect international mobility of capital and a fixed worldwide rate of
6 To emphasize the very real nature of psychological trauma as part of the distributional effect of a major policy shock to the economy, consider a 1993 article in the Washington Post. “In January 1991, after a bitter strike, Eastern Airlines grounded its planes forever. In a stroke, the 30,000 highly skilled and well-paid Eastern employees -- most of whom had 20 or 30 years with the company – joined the ranks of the jobless. Just 11 months later, Pan Am, the one-time aviation giant, went under. When its remaining 12,000 employees arrived at work on Dec. 4, 1991, security staff gave them one hour to clear out. A year and a half later, suicide among these laid off workers has reached epidemic proportions. Since Pan Am's demise, eight former employees have killed themselves – double the normal rate for men in their forties and fifties. Since the Eastern strike began in 1989, at least 14 former employees have killed themselves, as did the wife of an Eastern pilot. In one case, a mechanic also shot his children.” (Barbara Koeppel, “For Airline Workers the Crash Can Be Fatal,” Washington Post, Sunday, September 5, 1993).
-10-
return that investors must earn to be willing to provide capital at all – in other words, a
horizontal supply curve faced by firms in a small open economy. In that case, owners of
capital cannot bear any burden whatever from any domestic policy, whether it be a tax on
capital, a tax on carbon, or a mandated ceiling on the pollution per unit output. And if
workers are relatively less mobile between countries, then only they can bear burdens on
the sources side. Thus, these alternative extreme assumptions about mobility yield starkly
different conclusions about the relative burdens of environmental policy on workers or
investors (through a lower wage or rate of return).
This discussion points to the importance of empirical measurement of capital
mobility between jurisdictions. Moreover, that degree of mobility must also depend on
whether the jurisdictions are different provinces within a country, or different countries.
Severe legal constraints might hinder factor flows between countries with different policies
about the inflow of labor or capital. Or, the jurisdiction imposing the tax or other policy
under consideration might be a European country within the European Union, where labor
and capital mobility is greater, or the policy might be enacted by California only – with
much greater factor mobility between U.S. states.7
Further analysis in Gravelle and Smetters (2006) suggests two further points. First,
they show that imperfect substitution between imports and exports of the industry’s
outputs in an open-economy model can allow much of the burden to remain on capital
owners. Second, the burden on capital owners also depends on savings behavior, since
even a closed economy does not have a fixed vertical capital supply curve if savers
respond to lower rates of return by saving less. The bottom line, for now, is that the
possible burdens on capital from environmental policy remains an open question. And
while labor is less mobile internationally, workers also can change their labor supply in
response to changes in the real net wage, and so the extent of burden on labor also depends
on relative supply and demand elasticities in the labor market.
c. Inter-jurisdictional Trade Costs
Once we move beyond the model of a simple closed economy, the ease with which
products can be traded between jurisdictions is equally as relevant to the uses-side of
incidence as is factor mobility to the sources-side. If a substitute to a good produced and
7 Our focus here is on how openness affects factor returns and thus the distribution of burdens, but see Fowlie (2009) and others about how openness affects leakage – the increase in emissions in other jurisdictions.
-11-
consumed locally is available in a nearby jurisdiction with more lax regulations (or lower
tax rates), then consumers may substitute away from domestic purchases, and firms may
substitute away from domestic production. Analogously, if locations producing the same
goods vary in carbon intensity, a uniform tax on carbon production will place more burden
on producers in the more carbon-intensive location. On balance, the burden of a local tax
will shift away from the party less able to move their consumption or production to the
neighboring jurisdiction. For example, theory predicts that taxes levied on retail goods sold
near the border of a lower-tax jurisdiction are borne more heavily by producers, as
consumers can readily substitute away from local purchases in response to a tax increase.
Empirical research finds results broadly consistent with these predictions by using
inter-jurisdictional variation in tax rates. In energy, Doyle and Samphantharak (2008) and
Stolper (2016) estimate the pass-through of fuel taxes near borders at which tax rates
change discontinuously. In both studies, consumers near borders bear a lower fraction of
fuel tax burdens than do those far from the border, because these customers have more
elastic demand for the more-heavily-taxed item than do customers further from borders.8
d. Evasion and corruption
Classic models of incidence assume that regulations are enforced fully and that taxes are
collected costlessly. Yet, in many jurisdictions, evasion and corruption may prevent
jurisdictions from complete implementation of taxes or non-tax regulations. Here, we do
not draw sharp distinctions between tax avoidance, evasion, or bribing corrupt authorities
to relax enforcement, as all three cases may involve a choice to bear some lesser cost in
order to reduce the higher tax or regulatory cost. 9
Most existing literature on corruption and evasion focuses either on drivers of the
firm’s evasion decision10 or on fiscal effects of evasion and corruption – especially on
8 Outside of energy, a companion literature on cigarette tax pass-through finds similar patterns (Hanson and Sullivan (2009), Harding et al. (2012), DeCicca et al. (2013), and Chiou and Muehlegger (2014)). 9 In the normal parlance of public economics, tax avoidance is legal and tax evasion is not. However, either kind of activity may reduce taxes paid while incurring some additional cost. The taxpayer may incur extra fees paid to lawyers and accountants to reduce the legal tax burden. The consumer near a border may incur some extra cost of driving to the low-tax jurisdiction. Also, a bribe reflects a transfer of surplus between two parties, while avoidance or evasion may entail a real resource cost incurred to face less stringent regulation. 10 See Pomeranz (2015), Fisman and Wei (2004), de Paula and Scheinkman (2010), or Agostini and Martinez (2014).
-12-
government size, decentralization, and tax enforcement.11 A more modest literature
examines the relationship between evasion and the incidence of taxes (see e.g., Tanzi
(1992), Slemrod (2008), and Kopczuk et al. (2016)). Collectively, these papers highlight
how evasion and enforcement may influence the uses-side incidence of a tax or mandate.
Evasion reduces the effective rate of a tax or the effective stringency of an
environmental mandate. Parties engaged in evasion face the tax wedge as the benefit of
evasion, while incurring costs of the evasive activity. Evasion reduces tax revenues or the
efficacy of an environmental mandate. More subtly, tax evasion in part of a market can
impact tax incidence for fully-legal transactions in that market. Tax evasion reduces the
aggregate tax elasticity for the side of the market engaged in evasion. As a result, the
fully-legal party on the evasion-prone side of the market bears a higher burden of the tax
than in a world in which no evasion exists. Thus, if only a subset of firms evade a tax, the
burden of the tax on legal transactions falls more heavily on tax-compliant firms.
Intuitively, a tax increase makes tax-compliant firms less able to pass-through the extra tax
to consumers if they face competition from tax-evading firms.
Few papers map empirical evidence of evasion onto incidence. Kopczuk et al.
(2016) test the relationship between evasion and incidence by exploiting changes in the
identity of the party statutorily responsible to remit state fuel taxes. Consistent with
predictions, they find that the pass-through of diesel taxes to retail prices increased after
states moved the point of remittance from a point of the supply chain where monitoring
was difficult (retailers) to a point where enforcement was easier (wholesalers).
Although comprehensive indices of evasion and corruption are inherently imperfect
measures, the Global Competitiveness Report by the World Economic Forum generally
documents that survey-based measure of corruption or unethical firm behavior are
inversely related to country income.12 These statistics suggest that the impacts of evasion
and corruption might be especially relevant for developing countries.
e. Market Power
Finally, the standard model of incidence assumes perfect competition. For linear demand,
monopoly power reduces the pass-through of a tax onto consumers (because more of the
11 See, as examples, Sørensen (1994), Kau and Rubin (1981), Balke and Gardner (1991), or Gadenne and Singhal (2014). 12 Global Competitiveness Report, World Economic Forum, 2015-2016.
-13-
burden appears as a reduction in profits). In general, however, Bulow and Pfleiderer (1983)
show that the relationship between competition and pass-through is ambiguous, depending
on the curvature of supply and demand. Weyl and Fabinger (2013) extend this result to the
case of oligopoly, where the conduct of market participants can change with the imposition
of a tax. This research has renewed interest in the relationship between competition and
incidence, but even these new studies are in developed countries. Yet we still lack a clear
prediction about the effect of market concentration on incidence in developing countries.
Using partial equilibrium models, an empirical literature examines the relationship
between incidence and market power. Results are ambiguous, as causal identification is
complicated by unobservable factors correlated with the supply elasticity, demand
elasticity, and market concentration. Several papers examine market concentration and the
pass-through of diesel taxes. Doyle and Samphantharak (2008) find that gasoline tax pass-
through to retail customers is several percentage points higher in zip codes with more
concentrated market power and in zip codes with a smaller fraction of independent retail
stations. Kopczuk et al. (2016) also find weak evidence of higher pass-through of diesel
taxes in states with more concentrated wholesale markets.
A related literature estimates the pass-through of input cost shocks and, in contrast,
finds that competition tends to reduce pass-through rates. Miller et al. (2017) find evidence
of more than full pass-through of energy costs in the Portland cement industry, although
pass-through declines with proximity to rival firms. Ganapati et al. (2016) also find more
than full pass-through in the cement industry, but incomplete pass-through in five other
products including concrete and gasoline. Muehlegger and Sweeney (2017) study pass-
through of firm-specific and market-wide shocks in the refining industry and find that
pass-through declines when firms face local competition.
The bottom line is that the analyst’s job is not easy. The incidence of environmental
regulation depends on market concentration, the ease of evasion, trade costs, and mobility
of labor and capital between sectors or jurisdictions. And moreover, all of these market
conditions depend on the country being studied.
5. International Incidence of Carbon Policy
In order to look at incidence around the world, and to apply some of the concepts just
discussed, we now look at possible distributional effects of a climate policy such as a
carbon tax, a system of tradeable permits, or even command-and-control regulation. Any
-14-
such policy will raise the cost of fossil fuels and of goods produced using such fuels, so it
will impose burdens that depend on the relative carbon-intensity of each nation’s electric
power generation and of its exports. We therefore collect and discuss some relevant data
that can help determine whether more of this burden would likely be on rich or poor
countries. In order to focus on the distributional incidence of climate policy costs, we
ignore the benefits from reducing climate change, other potential co-benefits from reducing
local polluting emissions that are positively correlated with carbon emissions, and
distribution of benefits from the use of any carbon tax or permit revenues.13
Even with this focus on costs, however, the analysis could depend on many details
of the climate policy – and on who imposes it. A worldwide agreement to adopt a uniform
carbon tax seems unrealistic, of course, but one nation with a unilateral carbon tax places
itself at a competitive disadvantage abroad unless it also adopts border tax adjustments. It
could rebate the carbon tax already collected on production of goods that get exported; it
could impose a tax on the carbon embodied in goods imported from a country that does not
have a carbon tax; or it could do both.14 Incidence is complicated with only one such
adjustment, but both together effectively convert it from a tax on carbon used in production
to a tax on the carbon content of consumption. And since the tax would then apply to a
different activity, that policy choice affects the economic burden of a tax. For example, the
burden of a tax on carbon used in production could fall more on the owners of fossil fuel
resources and less on producers who can substitute towards renewable fuels, whereas the
burden of a tax on the carbon content of consumption goods can be passed from producers
in one country to consumers in other countries.
Because this paper is about distributional effects, we wish to avoid discussion of
how nations decide whether to impose unilateral carbon policy, or to join a coalition of
nations, or to impose border tax adjustments. Therefore, although we note the importance
13 A worldwide entity that levied a uniform carbon tax would impose burdens on countries with carbon- intensive production, as discussed here (ignoring benefits from the use of the revenue). But a country that participates in a worldwide agreement for all nations to impose a uniform carbon tax would generate revenue for itself. In this latter case, the carbon-intensive economy would not necessarily bear more burden as a nation, because it would generate more revenue from the tax that it could use for itself. Any actual worldwide agreement would likely be a mix of those outcomes, where rich countries pledge some of their carbon tax revenue to help poorer countries – in order to achieve their acquiescence to the agreement. 14 Here, we set aside the details of how one might calculate the increase in the final price of the imported good attributable to a carbon tax at each stage of production in the other country, or indeed whether the World Trade Organization would allow a tax upon import and a rebate upon export. For a more complete discussion of border tax adjustments, see Fischer and Fox (2012).
-15-
of such considerations, we now set aside further discussion of them. Instead, we discuss
the relative carbon-intensity of production in different nations around the world and the
incidence of a generic carbon tax that would raise the price of carbon-intensive products
and thus impose burdens on producers that depend on their supply elasticities and burdens
on consumers that depend on their demand elasticities. We interpret it as a worldwide tax,
simply so that a tax on carbon used in production is equivalent to a tax on the carbon
content of consumption: for a worldwide tax, border tax adjustments are irrelevant.
International trade allows for goods produced in two different countries to act as
substitutes, so any type of carbon policy can reduce demand for carbon-intensive products
from one country and increase demand for less-carbon-intensive products from elsewhere.
Thus, the distribution of burdens between countries is affected by carbon-intensity,
although the extent of these shifts in demand must depend on frictions such as imperfect
substitutability, transportation costs, and various protective national policies. In any case,
local factors of production used by firms in countries that rely on relatively carbon-
intensive sources of energy will bear more burden than those employed by firms with less
use of carbon. And, for this reason, production may shift toward a nation or sector that
avoids the worldwide tax (such as by poor enforcement capabilities).
These considerations suggest that a carbon policy is likely to have different
incidence for three categories of goods: (a) primary energy inputs such as coal, natural gas,
and crude oil; (b) energy-intensive produced commodities; and (c) non-energy-intensive
goods. For each group of commodities, one could examine both the incidence between
producers and consumers within a country (and effects on rich and poor who live there).
One could also examine the incidence between countries. Some data can shed light on how
these inter- and intra-jurisdictional effects are related to country income.15
a. Data
To address these questions, we collect data on country characteristics and trade patterns for
2013-2014. Data on country characteristics are published annually by the World Bank in
World Development Indicators (WDI).16 Most relevant for our analysis are GDP and GDP
15 Analogously, Mendelsohn et al. (2006) examine distributional impacts of climate change and find that its burdens fall heavily on poor countries. Others examine questions of environmental justice in the context of climate change (e.g., Ringius et al. (2002), Page (2008), and Lange et al. (2007)). 16 http://data.worldbank.org/data-catalog/world-development-indicators
-16-
per capita, electricity generation by fuel type, and industrial activity by sector. The World
Bank classifies countries into one of four income groups based on GDP per capita: low
income, lower-middle income, upper-middle income, and high income. We adopt their
classification as one method of grouping countries by income. Most of the nations
classified as “low income” are located in Sub-Saharan Africa. “Lower-middle income”
nations include South Asian and Southeast Asian nations (e.g., India, Indonesia, and
Vietnam). “Upper-middle income” nations includes China, Russia, Brazil, Turkey and
South Africa. “High income” nations are predominately from the OECD. To summarize
these categories, Appendix Table A1 presents country counts by income group and region
of the world.
We supplement the WDI data with country-level data on the value of imports and
exports by industry from the U.N. Commodity Trade Statistics Database (COMTRADE), a
comprehensive panel of country-to-country product flows.17 We aggregate export values
into energy-intensive goods and non-energy intensive goods, using industrial
classifications by the EIA as part of the International Energy Outlook.18 We separately
aggregate the value of fossil fuel exports, distinguishing four groups of fossil fuels: coal,
natural gas, crude oil, and other fuels (primarily refined petroleum products).
b. Fossil Fuel Production and Trade
We first examine primary production of fossil fuels to consider whether the burden of
carbon policy is likely to fall on producing or consuming states, using data from the
COMTRADE database to track exports and imports of fossil fuels.
Table 1 presents the mean and standard deviation of fossil fuel exports and imports
as a share of GDP. The first column presents statistics for all countries; columns two
through five present statistics by WDI income group. The bold value in the first row is the
total value of the four fossil fuel exports as a fraction of GDP (and its standard deviation).
The bold value in row six is the total value of the four fossil fuel imports as a fraction of
GDP for those countries (and its standard deviation). The remaining rows present exports
or imports of particular fossil fuels as a fraction of GDP.
17 http://comtrade.un.org/db/default.aspx 18 See https://www.eia.gov/forecasts/ieo/pdf/industrial.pdf. Major energy-intensive industries include petroleum refining; food manufacturing; pulp and paper; primary production of iron, steel, non-ferrous metals, and cement; and organic, agricultural and inorganic chemicals.
-17-
Across all nations, fossil fuel exports average 3.7% of GDP, of which roughly
eighty percent are crude and refined petroleum products (the crude oil average is 1.6% of
GDP, and refined products average 1.4% of GDP). Exports of natural gas are only 0.6% of
GDP, while coal exports are negligible (0.1% of GDP). Imports of all fuels constitute 4%
of GDP, where again petroleum products are greater than imports of coal and gas.
Worldwide averages in the first column of Table 1 obscure two sources of variation
across countries. First, on average, less wealthy countries are net importers of fossil fuels.
These imports in low-income countries average 8.1% of GDP. Exports for these countries
are modest, only 1.5% of GDP. Refined petroleum products (e.g., gasoline and diesel)
make up the vast majority of the imports for low-income countries, 7.6% of their GDP. In
contrast, wealthy countries both import and export fossil fuels; high income countries are
slight net importers, importing fossil fuels worth 4.0% of GDP and exporting fuel worth
3.4% of GDP.
Second, even within each income group, Table 1 shows substantial variation in
fossil fuel exports and imports.19 Fossil fuel exports are a high fraction of GDP for
resource-rich nation in the Middle-east, central Asia (e.g., Russia, Kazakhstan, and
Azerbaijan), and a handful of other locations (e.g., Norway, Bolivia, Brunei and Nigeria).
In seven countries, exports of natural gas exceed 5% of GDP: Qatar (42% of GDP), Brunei
(33%), Bolivia (19%), Algeria (11%), Norway (7.5%), Oman (6.8%) and Malaysia (6.5%).
Coal exports tend to be a smaller fraction of GDP; the five countries for which coal exports
are greatest relative to GDP are Mongolia (6.7%), Australia (2.2%), Indonesia (2.0%),
Columbia (1.6%) and South Africa (1.4%).
Trade patterns by income group beg the question of whether the burden of carbon
policy is likely to fall on consumers or producers of fossil fuels. As noted above, incidence
depends on the availability of substitutes for a taxed good, or more generally, on
elasticities of supply and demand. Thus, we separately consider the incidence of a carbon
policy on petroleum products (crude and refined products) and fossil fuels used for
electricity generation (coal or natural gas).
For refined petroleum products such as gasoline and diesel fuel, carbon content is
unrelated to where the fuel was produced. Fully-combusted, a gallon of gasoline generates
19 A hint of this variation can been seen in the column for “All Countries” by comparing the standard deviations of fossil fuel exports (0.076) and fossil fuel imports (0.036).
-18-
roughly 20 pounds of CO2, and a gallon of diesel fuel generates roughly 22 pounds of CO2.
We use as a benchmark the EPA’s 2016 estimate of the social cost of carbon in 2020 of
$42 per ton carbon dioxide.20 Then, a Pigouvian tax at that rate levied on the carbon in
gasoline and diesel fuel would increase prices by approximately 40-50 cents per gallon. A
number of papers on developed countries (e.g., Marion and Muehlegger (2011), or
Chouinard and Perloff (2004)) find that consumers bear the vast majority of gasoline taxes
and input cost changes in the short-run. Demand for transportation fuels is inelastic in the
short-run, allowing producers to shift the burden of taxes or input cost changes almost fully
onto consumers. If demand in developing countries is equally inelastic, then the burden of
a carbon policy on transportation fuels is likely borne by consumers (importing countries)
rather than producers (exporting countries).21
Long-run demand is more elastic, as consumers shift towards modes of
transportation that are less fuel intensive, such as vehicles with higher fuel-efficiency,
alternative fuel vehicles, or alternative methods of commuting (e.g., public transit). If these
shifts significantly increase the elasticity of demand for these fuels, the burden of carbon
policy might be shared more equally between consumers and producers. In practice,
though, a 40-50 cent jump in prices of transportation fuels, while significant, is of a
magnitude similar to other recent jumps in input prices that did not spur substantial shifts
towards alternative fuel vehicles. Although Li et al. (2014) for the U.S. and Rivers and
Schaufele (2015) for Canada find evidence that consumers respond more to taxes than to
changes in the input price of crude, even these estimates imply relatively modest tax-
induced shifts in fleet fuel economy. Busse et al. (2013) estimate that the new vehicle
market share of the most fuel-efficient quartile of vehicles rises about 3.5 percentage points
in response to a 50-cent per gallon increase in prices. Also for the U.S., Li et al. (2009)
estimate that a tax increase of comparable magnitude would increase average fuel economy
of purchased new vehicles by 1.1% and fleet-wide fuel economy by 0.13%. Gallagher and
Muehlegger (2011) estimate that an increase from $3.00 per gallon to $3.50 per gallon
leads to an 11% increase in sales of hybrid vehicles.
More research is needed on price sensitivity of consumers in developing
20 https://www.epa.gov/sites/production/files/2016-12/documents/sc_co2_tsd_august_2016.pdf 21 Though we discuss only a “generic” carbon tax, this statement about relative burdens holds generally, regardless of details such as how many countries have adopted policies that raise carbon prices.
-19-
countries.22 But if those consumers are similar to those in developed countries, the
implication is that even the long run burden of a carbon tax on fuels is likely to be felt
more by countries that are net importers of fuel.
Petroleum products have few easy substitutes in transportation, but natural gas and
coal compete more directly as inputs in the electric power sector and as thermal inputs for
industry.23 Cullen and Mansur (2017) infer the empirical effect of a carbon tax on the
substitution of coal-fired for gas-fired electricity generation by examining how electricity
generation changes in response to fluctuations in the commodity spot prices of the two
fuels. They find that a tax of $42 / ton CO2 has the potential to make coal-fired electricity
more expensive than electricity generated by combined cycle natural gas plants. Of course,
the dispatch order is a short run response, while the construction of new plants can shift
from coal to natural gas in the long run.
A carbon tax of $42 / ton CO2 is equivalent to a tax of $2.46 / mmBTU on natural
gas and of $4.43 / mmBTU on coal.24 Taking into account the greater efficiency of
combined cycle natural gas generators magnifies this advantage:25 a carbon tax of $42/ton
CO2 would increase the cost of burning coal by $5/mmBTU relative to natural gas.
The carbon intensity of coal relative to natural gas suggests that the short-run
burden of carbon policy falls more heavily on coal, as consumers can substitute towards
natural gas. With increases in the number and economic power of fossil-fuel-consuming
countries that adopt carbon policy, substitution away from coal and towards natural gas
will affect the worldwide market prices for coal and natural gas, thus imposing relative
burdens on coal producing and coal-dependent economies.
c. Carbon-intensity of Domestic Electricity Production
We now consider the impacts of carbon policy on electricity and manufactured goods. The
carbon-intensity of a manufactured good is the product of the energy-intensity at all stages
22 Sterner (2012) reviews some initial estimations of fuel price elasticities in developing countries. 23 Worldwide in 2015, the electricity power sector consumed approximately 60% of coal production and 34% of natural gas production. Source: EIA International Energy Outlook 2016, Table F-1 24 According to http://www.eia.gov/tools/faqs/, the carbon content of coal is 211 lbs/mmBTU, and of natural gas is 117 lbs/mmBTU for U.S. generators. If electricity generation is less efficient in developing countries, these numbers might understate the relative burden. 25 The average heat rate for U.S. coal and natural gas combined cycle generators were 10,080 and 7,658 mmBTU/kwh, respectively (http://www.eia.gov/electricity/annual/html/epa_08_02.html).
-20-
of production and carbon-intensity of the energy source. As a result, identical goods
produced in different countries may have differential carbon-intensity that depends on the
carbon-intensity of local energy production.
We construct a proxy for the weighted-average carbon-intensity of each country’s
local electricity generation by using its fractions of electricity generated from different fuel
sources (from the WDI) as weights for the estimated CO2 emissions per kwh for each fuel
source (from the EIA).26 In doing so, we make several implicit assumptions about the
nature of country-level energy production. First, the EIA’s estimates of CO2-per-kwh for
each fuel are based on data from U.S. generators. To the extent that electricity generation is
less efficient in developing countries, these numbers might understate carbon-intensity in
developing countries. It also abstracts away from an obvious shortcoming of analyzing
average costs in the electricity market: a carbon tax of sufficient magnitude may shift the
merit order of electricity plants, causing suppliers to substitute away from carbon-intensive
sources of electricity (such as coal) towards less carbon-intensive sources (like natural gas,
nuclear, or renewables). Substitution of this form would likely attenuate the absolute
competitive impact of carbon policy.27 Finally, the proxy only reflects the carbon-intensity
of electricity production; we do not observe direct industrial consumption of different
fossil fuels, so a full characterization of the carbon-intensity of all energy consumption is
beyond our capability here. If the mix of fossil fuels used directly by industry differs from
the mix used for electricity generation, we may overstate or understate the carbon-intensity
of industrial production.28
Table 2 summarizes the relative carbon intensity of domestic electricity production
across all countries in the World Bank data along with the sources of electricity. The first
row of Table 2 presents the mean carbon-intensity (and standard deviation) of electricity
generation – for all countries in the first column and by income group in columns two
through five. Worldwide, electricity generation averages 1.093 pounds of CO2 emissions
26 http://www.eia.gov/tools/faqs/. 27 Less likely are changes in relative positions of different countries because of substitution to different fossil sources for electricity generation. Carbon-intensive countries might shift production towards less-carbon-intensive fuels, but they are unlikely to leapfrog other countries that currently generate electricity in less carbon-intensive ways. Thus, conclusions about relative winners and losers would likely remain unchanged. 28 Although we cannot observe industrial consumption of fossil fuels directly, we can compare the use of fossil fuels for electricity generation and the export of those same fuels. The correlation between the fraction of generation and exports as a fraction of GDP for coal is 0.38, for oil is 0.43, and for natural gas is 0.30.
-21-
per kwh, but this carbon intensity is greatest for countries in the middle of the income
distribution (1.297 for lower-middle income countries and 1.217 for upper-middle income
countries). To explain these differences, other rows of Table 2 show fractions of electricity
generated from coal, oil, natural gas, nuclear, hydroelectricity, and other renewable
sources. The high carbon intensity in middle-income countries is attributable to their
relatively high generation of coal-fired electricity. In contrast, low-income countries
generate a greater fraction of electricity from hydropower and emit only 0.644 pounds of
CO2 per kwh of electricity. High-income countries generate a higher fraction of electricity
from nuclear and other renewable sources and produce 1.018 pounds of CO2 per kwh.
Figure 1: GDP per Capita and Average CO2 Emissions per KWH
To better convey the substantial variation in carbon-intensity of electricity
generation within each income group, Figure 1 plots each country’s carbon intensity (in
pounds of CO2 per kwh) against the log of GDP per capita. We highlight the world’s ten
largest economies in red. 29 In addition, we plot the quadratic line that best fits the data.
The inverted U-shaped shape of the quadratic line matches the pattern of average
29 See http://wits.worldbank.org/wits/wits/witshelp/content/codes/country_codes.htm for the three-letter abbreviations for each country.
AGO
ALB
ARE
ARG
ARM
AUS
AUT
AZE
BEL
BEN
BGD
BGR
BHRBIH
BOL
BRA
BRN
BWA
CAN
CHE
CHL
CHN
CIV
CMR COL
CRI
CYP
CZE DEU
DNK
DOM
DZA
ECU
EGY
ESP
ESTETH
FIN
FRA
GAB
GBR
GEOGHA
GRC
GTM
HKG
HND
HRV
HTI
HUN
IDNIND
IRL
IRN
ISL
ISR
ITA
JAMJOR
JPN
KAZ
KEN
KGZ
KHM
KOR
KWTLBN
LBY
LKA LTU
LUX
LVA
MAR
MDAMEX
MKD
MLT
MMR
MNE
MNG
MOZ
MUSMYS
NAM
NGA
NIC
NLD
NORNPL
NZL
OMN
PAK
PANPER
PHL
POL
PRT
PRY
QAT
RUS
SAU
SEN
SGP
SLV
SRB
SVK
SVN
SWE
THATTOTUN
TUR
TZA UKR
URY
USA
VEN
VNM
YEM
ZAF
ZMB
ZWE
BRA
CHN
DEU
FRA
GBR
IND
ITA
JPN
RUS
USA
0.5
11.
52
CO
2 em
issi
ons
per k
wh
6 8 10 12Log GDP per capita ($)
-22-
carbon-intensity in Table 2. To be clear, however, this curve should not be interpreted as
an “Environmental Kuznets Curve”. It is intended only to show a cross-sectional snapshot
of carbon intensity and not a causal relationship between income and emissions. Indeed,
the point of Figure 1 is the weakness of the relationship between income and emissions.
Roughly 95% of the variation in CO2 emissions per kwh is unexplained by GDP per capita.
In other words, countries at very similar levels of per-capita income vary tremendously
with respect to their carbon-intensity, which depends instead on the nature of local
electricity production.
In four of the five wealthiest OECD countries (U.S., Japan, Germany, France, and
Great Britain), electricity generation is modestly more carbon intensive than their average.
The outlier is France, which generates a high fraction of electricity from nuclear power and
is less carbon-intensive than average. Amongst “BRIC” countries (Brazil, Russia, India
and China), electricity generation in Brazil is less carbon-intensive because of its reliance
on hydropower. In contrast, China and India are more carbon-intensive, both generating
over seventy percent of their electricity from coal.
To put the vertical scale of Figure 1 in perspective, we can translate electricity
carbon intensity into dollar terms using the EPA’s 2016 estimate of the social cost of
carbon, $42 per ton of CO2. At this “price,” a country producing electricity that creates one
more pound of CO2 per kilowatt-hour (about half the height of Figure 1) would have
generation costs that are 2 cent per kilowatt-hour higher. As a point of reference, the
average retail electricity price in the U.S. in 2014 was roughly 10 cents per kilowatt-hour.
d. Trade in Energy-Intensive Goods
We now examine whether the carbon intensity of domestic energy production is correlated
with the goods traded by a country. Specifically, we examine whether carbon-intensive
nations also tend to export energy-intensive goods (such that they would face greater
exposure to climate policy). As a first step, we calculate COMTRADE exports of energy-
intensive goods as a percentage of GDP, and we find little correlation with income: the
percentage does not vary substantially between low-income countries (4.8%), lower-
middle-income (4.1%), upper-middle-income (3.8%), and high-income countries (4.7%).
Those calculations make it hard to suggest that trade in carbon-intensive goods introduces
-23-
much difference in burdens between rich and poor countries.
We further explore the relationship between carbon intensity of electricity and the
energy intensity of exports within each income group. For each of the four income groups,
Figure 2 plots the carbon-intensity of a country’s electricity generation against its exports
of energy-intensive goods as a fraction of GDP.
Each panel includes a horizontal dotted line corresponding to worldwide average
carbon-intensity of electricity (0.99 lbs of CO2 per kwh), and a vertical line corresponding
to worldwide average ratio of energy-intensive exports to GDP (4.4%). Thus, any country
at the top of a panel has relatively carbon-intensive electricity, and any to the right exports
a higher fraction of energy-intensive goods (has more trade vulnerability).
Figure 2: CO2 Emissions per kwh and Energy-intensive exports
by Income Groups
Some combination of these two measures can indicate the degree to which a
country’s exports might determine burden from carbon policy.30 Thus, countries whose
30 Differential carbon-intensity of imports matters less: countries that import energy-intensive goods can substitute towards less carbon-intensive sources for the same goods. Easier substitution pushes more burden onto carbon-intensive exporters (which is why we are looking at carbon-intensity of exporters).
BEN
ETH MOZ
NER
NPL
SEN
TGO
TZA ZWE
ARM
BOLCIV
CMRCOG
EGY
GTM
HND
IDNIND
KHMLKA
MAR
MDA
MNG
NGANIC
PAK
PHL
SLV
UKRVNM
YEM
ZMB
IND
ALB
AZEBGR
BIHBLR
BRA
CHN
COL
DOMDZA
ECU
GEO
IRQ
JAMJORKAZLBN
MEX
MKD
MNE
MUSMYS
NAM
PAN PER
PRY
ROMRUS
SRBTHA
TUR
ZAF
BRA
CHN
RUS
AREARG
AUS
AUT BEL
BHRBRN
CAN
CHE
CHL
CYP
CZEDEUDNK
ESP
EST
FIN
FRA
GBRGRC
HRV HUN
IRL
ISL
ISR
ITA
JPN KOR
KWT
LTULUX
LVA
MLT
NLD
NOR
NZL
OMN
POL
PRT
QATSAU
SVK
SVN
SWE
URY
USA DEU
FRA
GBRITA
JPNUSA
0.5
11.
52
0.5
11.
52
0 .1 .2 .3 0 .1 .2 .3
Low income Lower middle income
Upper middle income High income
CO
2 Em
issi
ons
per K
WH
Energy-intensive non-fuel exports (% of GDP)Horizontal and vertical lines correspond to the median values of CO2 Emissions per KWH(0.99 lbs/kwh) and non-fuel, energy-intensive exports (5% of GDP)
-24-
exports are most exposed to climate policy are those in the upper-right hand quadrant, such
as Poland (POL), which exports a high-fraction of energy-intensive goods using relatively
carbon-intensive sources of production. In contrast, countries in the lower-right hand
quadrant, such as Iceland (ISL), export a similarly high fraction of energy-intensive goods
as a fraction of GDP, but do so using less-carbon intensive energy sources.
Thus, if enough countries institute carbon policies that lead to reduced demands for
carbon-intensive goods, then countries above the horizontal dotted line (at 0.99 lbs / kwh)
would be placed at a competitive disadvantage (and thus face more burden). Countries like
Iceland that generate electricity through low-carbon means while exporting energy-
intensive goods could face a negative burden – a gain from worldwide carbon policy.
When grouped by income in Figure 2, an interesting pattern emerges. The
relationship between exports and carbon intensity varies substantially by income group. In
particular, upper-middle-income countries like South Africa (ZAF) tend both to produce
carbon-intensive electricity and to export a high fraction of energy-intensive goods. These
are countries most likely to be placed at a competitive disadvantage in a world of unified or
at least widespread carbon policy. In contrast, high-income countries are more likely to fall
in the lower-right hand quadrant and are thus likely to receive a competitive advantage in a
world of widespread carbon policy. Despite variation in carbon-intensity, the ten largest
countries in the world (in red) are located to the left in the figure, which means they are
less exposed to changes in competitiveness of domestic energy-intensive manufacturing,
by virtue of their size and diversification of their economies.
Within a country, carbon policy may also affect the distribution of economic
activity across energy-intensive and non-energy intensive goods. Carbon policy raises the
price of energy-intensive goods relative to non-energy-intensive goods, so it likely leads to
a shift away from production of energy-intensive goods toward production of non-energy-
intensive goods. It therefore implies short-run burdens on sector-specific factors, as the tax
is capitalized into their market value. In the long run, the mix of industries in countries
with carbon-intensive energy will shift towards less trade-exposed industries, with the exit
of local companies in trade-exposed, energy-intensive industries.
6. Conclusion
-25-
In this paper, we consider the distributional consequences of environmental regulations,
either taxes or mandates, focusing specifically on how lessons and estimates from the
developed world might map to the international context. Although mandates account for
the majority of environmental regulations internationally, the literature focuses heavily on
environmental taxes, and largely in the context of developed countries. Yet, results from
the U.S. and Europe are not directly applicable to the developing world, where future
environmental regulations may be most relevant, yet where market frictions and
imperfections may be more pronounced.
We summarize the literature related to taxes and other market-based approaches,
arguing that analogous logic can apply to the case of environmental mandates. We then
summarize how deviations from the initial simple model affect incidence on the sources-
side and uses-side. Finally, we gather country-level data on carbon emissions, trade, and
per capita income, and we consider how policies to address carbon emissions might
impose distributional burdens across countries and within countries, in both the developed
and developing world.
References Agostini, Claudio A., and Claudia Martinez. 2014. Tax credits response to tax
enforcement: Evidence from a quasi-experiment in Chile. Fiscal Studies 35 (1): 41-65. Balke, Nathan S., and Grant W. Gardner. 1991. Tax collection costs and the size of
government. Working paper, Mimeo, Southern Methodist University. Bruegge, Chris, Tatyana Deryugina, and Erica Myers. 2016. The distributional effect of
building codes. Working paper, University of Illinois at Champaign-Urbana. Bulow, Jeremy I. and Paul Pfleiderer. 1983. A note on the effect of cost changes on
prices. Journal of Political Economy 91 (1): 182-185. Bushnell, James B., Howard Chong, and Erin T. Mansur. 2013. Profiting from regulation:
Evidence from the European carbon market. American Economic Journal: Economic Policy 5 (4): 78–106.
Busse, Meghan R., Christopher R. Knittel, and Florian Zettelmeyer. 2013. Are consumers
myopic? Evidence from new and used car purchases. The American Economic Review 103 (1): 220-256.
-26-
Chiou, Lesley, and Erich Muehlegger. 2014. Consumer response to cigarette excise tax changes. National Tax Journal 67 (3): 621-650.
Chouinard, Hayley, and Jeffrey M. Perloff. 2004. Incidence of federal and state gasoline
taxes. Economics Letters 83 (1): 55-60. Cronin, Julie Anne, Don Fullerton, and Steven Sexton. 2017. Vertical and horizontal
redistribution from a carbon tax and rebate. NBER Working Paper No. 23250. Cullen, Joseph A., and Erin T. Mansur. 2017. Inferring carbon abatement costs in
electricity markets: A revealed preference approach using the shale revolution. American Economic Journal: Economic Policy (forthcoming).
Davis, Lucas, and Christopher R. Knittel. 2016. Are fuel economy standards regressive?
NBER Working Paper No. 22925, NBER, Cambridge, MA. de Paula, Áureo, and Jose A. Scheinkman. 2010. Value-added taxes, chain effects and
informality. American Economic Journal: Macroeconomics 2 (4): 195-221. DeCicca, Philip, Donald Kenkel, and Feng Liu. 2013. Who pays cigarette taxes? The
impact of consumer price search. Review of Economics and Statistics 95 (2): 516-529. Doyle, Joseph J., and Krislert Samphantharak. 2008. $2.00 Gas! Studying the effects of a
gas tax moratorium. Journal of Public Economics 92 (3): 869-884. Fischer, Carolyn, and Alan K. Fox. 2012. Comparing policies to combat emissions
leakage: Border tax adjustments versus rebates. Journal of Environmental Economics and Management 64 (2):199-216.
Fisman, Raymond, and Shang-Jin Wei. 2004. Tax rates and tax evasion: Evidence from
missing imports in China. Journal of Political Economy 112 (2): 471-496. Fowlie, Meredith L. 2009. Incomplete environmental regulation, imperfect competition,
and emissions leakage. American Economic Journal: Economic Policy 1 (2): 72-112. Fullerton, Don, and Garth Heutel. 2007. The general equilibrium incidence of
environmental taxes. Journal of Public Economics 91 (3-4): 571-591. Fullerton, Don, and Garth Heutel. 2010. The general equilibrium incidence of
environmental mandates. American Economic Journal: Economic Policy 2 (3): 64-89. Fullerton, Don, and Gilbert Metcalf. 2002. Tax incidence. In Handbook of public
economics, eds. A. Auerbach, and M. Feldstein. Amsterdam: Elsevier. Gadenne, Lucie, and Monica Singhal. 2014. Decentralization in developing economies.
Annual Review of Economics 6 (1): 581-604. Gallagher, Kelly Sims, and Erich Muehlegger. 2011. Giving green to get green? Incentives
and consumer adoption of hybrid vehicle technology. Journal of Environmental
-27-
Economics and Management 61 (1): 1-15. Ganapati, S., Joseph S. Shapiro, and Reed Walker. 2016. Energy prices, pass-through, and
incidence in US manufacturing. NBER working paper No. 22281, Cambridge MA. Gibson, Matthew. 2015. Regulation-induced pollution substitution. Working paper,
Department of Economics, William College. Goulder, Lawrence H., Marc A. C. Hafstead, and Roberton C. William III. 2016. General
equilibrium impacts of a federal clean energy standard. American Economics Journal: Economics Policy 8 (2): 186-218.
Gravelle, Jane G., and Kent A. Smetters. 2006. Does the open economy assumption really
mean that labor bears the burden of a capital income tax? The B.E. Journal of Economic Analysis & Policy 6 (1): 1-42.
Greenstone, Michael, and Rema Hanna. 2014. Environmental regulations, air and water
pollution and infant mortality in India. American Economic Review 104 (10): 3038-3072.
Greenstone, Michael, and B. Kelsey Jack. 2015. Envirodevonomics: A research agenda for
an emerging field. Journal of Economic Literature 53 (1): 5-42. Hanson, Andrew, and Ryan Sullivan. 2009. The incidence of tobacco taxation: Evidence
from geo-graphic micro-level data. National Tax Journal 62 (4): 677-698. Harberger, Arnold C. 1962. The incidence of the corporation income tax. Journal of
Political Economy 70 (3): 215-240. Harding, Matthew, Ephraim Leibtag, and Michael F. Lovenheim. 2012. The heterogeneous
geo-graphic and socioeconomic incidence of cigarette taxes: Evidence from Nielsen Homescan data. American Economic Journal: Economic Policy 4 (4): 169-198.
Kau, James B., and Paul H. Rubin. 1981. The size of government. Public Choice 37 (2):
261-274. Kopczuk, Wojciech, Justin Marion, Erich Muehlegger, and Joel Slemrod. 2016. Does tax-
collection invariance hold? Evasion and the pass-through of state diesel taxes. American Economic Journal: Economic Policy 8 (2): 251-286.
Lange, Andreas, Carsten Vogt, and Andreas Ziegler. 2007. On the importance of equity in
international climate policy: An empirical analysis. Energy Economics 29 (3): 545-562. Levinson, Arik. 2016. Energy efficiency standards are more regressive than energy taxes:
Theory and evidence. NBER Working Paper No. 22956, NBER, Cambridge, MA. Li, Shanjun, Joshua Linn, and Erich Muehlegger. 2014. Gasoline taxes and consumer
behavior. American Economic Journal: Economic Policy 6 (4): 302-342.
-28-
Li, Shanjun, Christopher Timmins, and Roger H. von Haefen. 2009. How do gasoline prices affect fleet fuel economy? American Economic Journal: Economic Policy 1 (2): 113-137.
Marion, Justin, and Erich Muehlegger. 2011. Fuel tax incidence and supply conditions.
Journal of Public Economics 95 (9): 1202-1212. Mendelsohn, Robert, Ariel Dinar, and Larry Williams. 2006. The distributional impact of
climate change on rich and poor countries. Environment and Development Economics 11 (2): 159-178.
Mieszkowski, Peter. 1972. The property tax: An excise tax or a profits tax? Journal of
Public Economics 1 (1): 73-96. Miller, Nathan H., Matthew Osborne, and Gloria Sheu. 2017. Pass-through in a
concentrated industry: Empirical evidence and regulatory implications. The RAND Journal of Economics 48 (1): 69–93.
Muehlegger, Erich and Richard Sweeney. 2017. Competition and pass-through of input
cost shocks: Evidence from the oil shale boom. Working paper, U.C. Davis, CA. Page, Edward A. 2008. Distributing the burdens of climate change. Environmental Politics
17 (4): 556-575. Pomeranz, Dina. 2015. No taxation without information: Deterrence and self-enforcement
in the value added tax. American Economic Review 105 (8): 2539-2569. Reguant, Mar. 2017. The efficiency and distributional impacts of large-scale renewable
policies. Working paper, Northwestern University. Ringius, Lasse, Asbjørn Torvanger, and Arild Underdal. 2002. Burden sharing and fairness
principles in international climate policy. International Environmental Agreements 2 (1): 1-22.
Rivers, Nicholas, and Brandon Schaufele. 2015. Salience of carbon taxes in the gasoline
market. Journal of Environmental Economics and Management 74: 23-36 Slemrod, Joel. 2008. Does it matter who writes the check to the government? The
economics of tax remittance. National Tax Journal 61 (2): 251-275. Sørensen, Peter Birch. 1994. From the global income tax to the dual income tax: Recent
tax reforms in the Nordic countries. International Tax and Public Finance 1 (1): 57-79. Sterner, Thomas (Editor). 2012. Fuel Taxes and the poor: The distributional effects of
gasoline taxation and their implications for climate policy. Washington DC: RFF Press.
Stolper, Samuel. 2017. Competition and incidence: Automotive fuel tax pass-through at
state borders. MIT working paper, Cambridge MA.
-29-
Sullivan, Daniel, and Till Von Wachter. 2009. Job displacement and mortality: An analysis
using administrative data. The Quarterly Journal of Economics 124 (3): 1265-1306. Tanzi, Vito. 1992. Theory and policy: A comment on Dixit and on current theory. Staff
Papers- International Monetary Fund 39 (4): 957-966. Weyl, E. Glen, and Michal Fabinger. 2013. Pass-through as an economic tool: Principles
of incidence under imperfect competition. Journal of Political Economy 121 (3): 528-583.
-30-
Table 1: Imports and Exports of Fossil Fuels as Fractions of GDP, by income group
All
countries Low income
countries Lower middle
income Upper middle
income High
income All Fossil Fuel Exports 0.037 0.015 0.050 0.043 0.034 (0.076) (0.021) (0.052) (0.074) (0.078) Coal Exports 0.001 0.000 0.004 0.001 0.001 (0.004) (0.000) (0.008) (0.003) (0.004) Natural Gas Exports 0.006 0.002 0.007 0.006 0.005 (0.026) (0.008) (0.018) (0.016) (0.029) Crude Oil Exports 0.016 0.000 0.021 0.022 0.013 (0.054) (0.000) (0.046) (0.053) (0.054) Refined Fuel Exports 0.014 0.013 0.017 0.013 0.014 (0.023) (0.015) (0.014) (0.020) (0.025) All Fossil Fuel imports 0.040 0.081 0.066 0.034 0.040 (0.036) (0.028) (0.029) (0.026) (0.039) Coal Imports 0.001 0.001 0.004 0.001 0.001 (0.002) (0.001) (0.004) (0.001) (0.002) Natural Gas Imports 0.006 0.003 0.007 0.004 0.007 (0.007) (0.004) (0.007) (0.005) (0.007) Crude Oil Imports 0.020 0.002 0.035 0.017 0.020 (0.018) (0.008) (0.031) (0.016) (0.016) Refined Fuel Imports 0.013 0.076 0.020 0.012 0.012 (0.021) (0.026) (0.017) (0.018) (0.022)
Source: COMTRADE data, United Nations. Note: For each fuel type, the cell denotes fossil fuel exports as a fraction of GDP in 2013 (and standard deviation). In low-income countries, for example, total fossil fuel exports averaged 1.5% of GDP (for the fuel types aggregated by value), while total imports of fossil fuels averaged 8.1% of GDP (aggregated by value). All group statistics are weighted by country GDP.
-31-
Table 2: Carbon intensity of domestic electricity production, by income group
All countries
Low income countries
Lower middle income
Upper middle income
High income countries
CO2 emissions per kwh 1.093 0.644 1.297 1.217 1.018 (0.441) (0.695) (0.383) (0.474) (0.409) Coal Generation (% of Total) 35.189 4.531 39.379 42.856 31.565 (24.723) (14.213) (30.678) (33.587) (17.575) Oil Generation (% of Total) 3.808 24.654 8.444 4.359 2.950 (8.040) (35.740) (11.061) (9.629) (5.915) NG Generation (% of Total) 24.373 9.526 27.292 20.759 25.668 (20.282) (18.245) (28.663) (24.890) (16.331) Nuke Generation (% of Total) 12.541 0.000 2.683 3.657 17.531 (16.286) (0.000) (7.748) (5.105) (17.942) Hydro Generation (% of Total) 15.798 59.858 17.846 24.466 11.632 (19.460) (40.140) (17.392) (21.265) (17.045) Other Renew Gen (% of Total) 7.160 1.317 4.098 3.397 9.151
(6.772) (1.668) (4.871) (2.615) (7.354) Source: World Economic Indicators, World Bank. Note: The first row shows average carbon intensity (and standard deviation). For example, low-income countries average 0.644 pounds of CO2 emissions per kwh. Other rows show the mean and standard deviation of the fraction of electricity generation in 2013 from each different source. All group statistics are weighted by country GDP.
-32-
Appendix
Appendix Table A1: Countries by Region and Income Group, 2013
Low income
Lower middle income
Upper middle income
High income Total
East Asia & Pacific 1 15 8 13 37
Europe & Central Asia 0 7 14 37 58
Latin America & Caribbean 1 5 19 16 41
Middle East & North Africa 0 7 6 8 21
North America 0 0 0 3 3
South Asia 2 5 1 0 8
Sub-Saharan Africa 27 13 7 1 48
Total 31 52 55 78 216